Applying the brain age algorithm to the Insight 46 cohort study
Aaron Wagen
6/04/2020
1 Setup
Libraries
library(ggExtra)
library(kableExtra)
library(psych)
library(qwraps2)
library(stats)
library(tidyverse)
library(ggpubr)
library(broom)
library(skimr)
library(jtools)
require(ggforce)Set options and theme
Import data
mydata <- read.csv("brainagedatacleaned20200819.csv")Organise data
Relevel factors
mydata$SocioStatus <- factor(mydata$SocioStatus, levels = c("Manual", "Non-manual"))
mydata$Smoking <- factor(mydata$Smoking, levels = c("Never smoked", "Ex-smoker", "Current smoker"))
mydata$ApoE4 <- factor(mydata$ApoE4, levels = c("Non-carrier", "Carrier"))
mydata$Education <- factor(mydata$Education, levels = c("None attempted", "Vocational or GCSE", "A-level or higher"))
mydata$Diagnosis <- factor(mydata$Diagnosis)Recode the snp as a factor
mydata$rs1452628_T <- as.factor(mydata$rs1452628_T)Create dataset of all cases
mydata <- mydata %>%
dplyr::select(-id)
allcases <- mydata %>% #Create dataset of all cases regardless of missing data-points, select variables
dplyr::select(-Diagnosis, -Age_at_blood, -ScanAge)2 Missing data
Create the dataset of the sample used in the study (mydata5)
dim(mydata) #Remove 34 participants with missing brain predicted age values## [1] 502 31mydata2 <- mydata %>%
drop_na(BPA)
dim(mydata2)## [1] 468 31mydata3 <- mydata2 %>% #Remove 8 participants with missing amyloid status
drop_na(AmyloidStatus)
dim(mydata3)## [1] 460 31mydata4 <- mydata3 %>% #Remove 2 participants with missing ApoE4 status
drop_na(ApoE4)
dim(mydata4)## [1] 458 31mydata5 <- mydata4 %>% #Remove 2 participants with missing Serum NFL status
drop_na(SerumNFL)
dim(mydata5)## [1] 456 31mydata5 %>% #Check missingness in the core sample used in the study.
summarise_all(funs(sum(is.na(.)))) %>%
pivot_longer(everything(), names_to="Variable", values_to="NA Count") %>%
kable()| Variable | NA Count |
|---|---|
| Sex | 0 |
| SocioStatus | 0 |
| Education | 0 |
| Childcog | 0 |
| Smoking | 0 |
| MajorBrainDisorder | 0 |
| Diagnosis | 0 |
| Age | 0 |
| ScanAge | 0 |
| BPA | 0 |
| PAD | 0 |
| TIV | 0 |
| WBV | 0 |
| Ventvol | 0 |
| Hippovol | 0 |
| WMHVol | 0 |
| PACC | 0 |
| AmyloidStatus | 0 |
| AmyloidSuvr | 0 |
| ApoE4 | 0 |
| PRS | 30 |
| rs1452628_T | 33 |
| Age_at_blood | 0 |
| SerumNFL | 0 |
| Tauplasma | 3 |
| FHS_36 | 45 |
| FHS_69 | 13 |
| FEV1 | 47 |
| GripStrength | 7 |
| WalkSpeed | 25 |
| Height | 0 |
Create dataset of participants with complete data in all variables.
Scale continuous variables to facilitate comparisons
mydata5 <- mydata5 %>%
mutate(across(c(Age, PAD, WBV, Ventvol, TIV, WMHVol, Hippovol, AmyloidSuvr, SerumNFL, FHS_36, FHS_69, PRS, FEV1, GripStrength, WalkSpeed, Height), scale, .names = "{col}z"))3 Summary Statistics
Create summary table by total and grouped by sex using qwraps package
options(qwraps2_markup = "markdown", digits = 2)
summary2 <- #Summary used for table 1 without sex as a variable
list("Chronological age (years)" =
list("Range" = ~ paste(frmt(range(.data$Age, na.rm=T), digits=1), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Age, na_rm=T, digits=1, show_n = "never", denote_sd = "paren")),
"Brain predicted age (years)" =
list("Range" = ~ paste(frmt(range(.data$BPA, na.rm=T), digits=1), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$BPA, na_rm=T, digits=1, show_n = "never", denote_sd = "paren")),
"Brain predicted age difference (years)" =
list("Range" = ~ paste(frmt(range(.data$PAD, na.rm=T), digits=1), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PAD, na_rm=T, digits=1, show_n = "never", denote_sd = "paren")),
"Socioeconomic status" =
list("Manual" = ~ qwraps2::n_perc0(.data$SocioStatus == "Manual", na_rm = T, show_symbol = T),
"Non-manual"= ~ qwraps2::n_perc0(.data$SocioStatus == "Non-manual", na_rm = T, show_symbol = T)),
"Educational attainment" =
list("None attempted" = ~ qwraps2::n_perc0(.data$Education == "None attempted", na_rm = T, show_symbol = T),
"School grade" = ~ qwraps2::n_perc0(.data$Education == "Vocational or GCSE", na_rm = T, show_symbol = T),
"Higher education" = ~ qwraps2::n_perc0(.data$Education == "A-level or higher", na_rm = T, show_symbol = T)),
"Childhood cognition (z-score)" =
list("Range" = ~ paste(frmt(range(.data$Childcog, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Childcog, na_rm=T, show_n = "never", denote_sd = "paren")),
"Smoking" =
list("Never smoked" = ~ qwraps2::n_perc0(.data$Smoking == "Never smoked", na_rm = T, show_symbol = T),
"Ex-smoker" = ~ qwraps2::n_perc0(.data$Smoking == "Ex-smoker", na_rm = T, show_symbol = T),
"Current smoker" = ~ qwraps2::n_perc0(.data$Smoking == "Current smoker", na_rm = T, show_symbol = T)),
"Major brain disorder" =
list("None" = ~ qwraps2::n_perc0(.data$MajorBrainDisorder == "None", na_rm = T, show_symbol = T),
"Present"= ~ qwraps2::n_perc0(.data$MajorBrainDisorder == "Present", na_rm = T, show_symbol = T)),
"Total intracranial volume (ml)" =
list("Range" = ~ paste(frmt(range(.data$TIV, na.rm=T), digits=0), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$TIV, na_rm=T, show_n = "never", denote_sd = "paren", digits=0)),
"Whole brain volume (ml)" =
list("Range" = ~ paste(frmt(range(.data$WBV, na.rm=T), digits=0), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WBV, na_rm=T, show_n = "never", denote_sd = "paren", digits=0)),
"Ventricular volume (ml)" =
list("Range" = ~ paste(frmt(range(.data$Ventvol, na.rm=T), digits=2), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Ventvol, na_rm=T, show_n = "never", denote_sd = "paren", digits=2)),
"Hippocampal volume (ml)" =
list("Range" = ~ paste(frmt(range(.data$Hippovol, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Hippovol, na_rm=T, show_n = "never", denote_sd = "paren")),
"White matter hyperintensity volume (ml)" =
list("Range" = ~ paste(frmt(range(.data$WMHVol, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WMHVol, na_rm=T, show_n = "never", denote_sd = "paren")),
"PACC score (z-score)" =
list("Number" = ~ length(na.omit(.data$PACC)),
"Range" = ~ paste(frmt(range(.data$PACC, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PACC, na_rm=T, show_n = "never", denote_sd = "paren")),
"Amyloid status" =
list("Negative" = ~ qwraps2::n_perc0(.data$AmyloidStatus == "Negative", na_rm = T, show_symbol = T),
"Positive"= ~ qwraps2::n_perc0(.data$AmyloidStatus == "Positive", na_rm = T, show_symbol = T)),
"Amyloid SUVR" =
list("Range" = ~ paste(frmt(range(.data$AmyloidSuvr, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$AmyloidSuvr, na_rm=T, show_n = "never", denote_sd = "paren")),
"Serum neurofilament light (pg/ml)" =
list("Range" = ~ paste(frmt(range(.data$SerumNFL, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$SerumNFL, na_rm=T, show_n = "never", denote_sd = "paren")),
"APOE ε4 status" =
list("Non-carrier"= ~ qwraps2::n_perc0(.data$ApoE4 == "Non-carrier", na_rm = T, show_symbol = T),
"Carrier" = ~ qwraps2::n_perc0(.data$ApoE4 == "Carrier", na_rm = T, show_symbol = T)),
"Alzheimer's Polygenic Risk Score (z-score)" =
list("Number" = ~ length(na.omit(.data$PRS)),
"Range" = ~ paste(frmt(range(.data$PRS, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PRS, na_rm=T, show_n = "never", denote_sd = "paren")),
"rs1452628_T Allele number" =
list("No copies"= ~ qwraps2::n_perc0(.data$rs1452628_T == "0", na_rm = T, show_symbol = T),
"1 copy" = ~ qwraps2::n_perc0(.data$rs1452628_T == "1", na_rm = T, show_symbol = T),
"2 copies" = ~ qwraps2::n_perc0(.data$rs1452628_T == "2", na_rm = T, show_symbol = T)),
"Framingham Risk Score age 36" =
list("Number" = ~ length(na.omit(.data$FHS_36)),
"Range" = ~ paste(frmt(range(.data$FHS_36, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FHS_36, na_rm=T, show_n = "never", denote_sd = "paren")),
"Framingham Risk Score age 69" =
list("Number" = ~ length(na.omit(.data$FHS_69)),
"Range" = ~ paste(frmt(range(.data$FHS_69, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FHS_69, na_rm=T, show_n = "never", denote_sd = "paren")),
"Forced expiratory volume (L)" =
list("Number" = ~ length(na.omit(.data$FEV1)),
"Range" = ~ paste(frmt(range(.data$FEV1, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FEV1, na_rm=T, show_n = "never", denote_sd = "paren")),
"Grip strength (kg)" =
list("Number" = ~ length(na.omit(.data$GripStrength)),
"Range" = ~ paste(frmt(range(.data$GripStrength, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$GripStrength, na_rm=T, show_n = "never", denote_sd = "paren")),
"Walking speed (m/s)" =
list("Number" = ~ length(na.omit(.data$WalkSpeed)),
"Range" = ~ paste(frmt(range(.data$WalkSpeed, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WalkSpeed, na_rm=T, show_n = "never", denote_sd = "paren")))by_sex <- summary_table(dplyr::group_by(mydata5, Sex), summary2)
totalbysex <- summary_table(mydata5, summary2)
table1 <- cbind(totalbysex, by_sex)
print(table1,
rtitle = "Summary Statistics",
cnames = c("Total (N=456)", "Female (N=225)", "Male (N=231)"))##
##
## |Summary Statistics |Total (N=456) |Female (N=225) |Male (N=231) |
## |:----------------------------------------------|:-------------|:--------------|:-------------|
## |**Chronological age (years)** | | | |
## | Range |69.2, 71.9 |69.3, 71.9 |69.2, 71.8 |
## | Mean (SD) |70.7 (0.7) |70.7 (0.7) |70.7 (0.7) |
## |**Brain predicted age (years)** | | | |
## | Range |46.3, 94.3 |46.3, 85.2 |50.9, 94.3 |
## | Mean (SD) |67.9 (8.1) |65.2 (7.4) |70.6 (7.9) |
## |**Brain predicted age difference (years)** | | | |
## | Range |-24.6, 22.7 |-24.6, 14.7 |-19.9, 22.7 |
## | Mean (SD) |-2.8 (8.0) |-5.5 (7.3) |-0.1 (7.8) |
## |**Socioeconomic status** | | | |
## | Manual |70 (15%) |30 (13%) |40 (17%) |
## | Non-manual |386 (85%) |195 (87%) |191 (83%) |
## |**Educational attainment** | | | |
## | None attempted |70 (15%) |37 (16%) |33 (14%) |
## | School grade |139 (30%) |82 (36%) |57 (25%) |
## | Higher education |247 (54%) |106 (47%) |141 (61%) |
## |**Childhood cognition (z-score)** | | | |
## | Range |-1.60, 2.50 |-1.59, 2.47 |-1.60, 2.50 |
## | Mean (SD) |0.41 (0.75) |0.44 (0.74) |0.38 (0.75) |
## |**Smoking** | | | |
## | Never smoked |160 (35%) |86 (38%) |74 (32%) |
## | Ex-smoker |280 (61%) |131 (58%) |149 (65%) |
## | Current smoker |16 (4%) |8 (4%) |8 (3%) |
## |**Major brain disorder** | | | |
## | None |415 (91%) |207 (92%) |208 (90%) |
## | Present |41 (9%) |18 (8%) |23 (10%) |
## |**Total intracranial volume (ml)** | | | |
## | Range |1,114, 1,939 |1,114, 1,558 |1,274, 1,939 |
## | Mean (SD) |1,431 (133) |1,341 (92) |1,518 (106) |
## |**Whole brain volume (ml)** | | | |
## | Range |819, 1,494 |819, 1,265 |946, 1,494 |
## | Mean (SD) |1,099 (99) |1,045 (82) |1,151 (86) |
## |**Ventricular volume (ml)** | | | |
## | Range |6.16, 112.00 |6.16, 82.93 |9.33, 112.00 |
## | Mean (SD) |30.94 (16.34) |26.43 (14.70) |35.32 (16.68) |
## |**Hippocampal volume (ml)** | | | |
## | Range |4.12, 8.54 |4.12, 7.45 |4.83, 8.54 |
## | Mean (SD) |6.26 (0.67) |6.01 (0.59) |6.51 (0.65) |
## |**White matter hyperintensity volume (ml)** | | | |
## | Range |0.27, 33.67 |0.35, 32.78 |0.27, 33.67 |
## | Mean (SD) |5.21 (5.54) |5.64 (5.90) |4.80 (5.15) |
## |**PACC score (z-score)** | | | |
## | Number |456 |225 |231 |
## | Range |-3.49, 1.72 |-3.48, 1.67 |-3.49, 1.72 |
## | Mean (SD) |-0.01 (0.74) |0.16 (0.73) |-0.17 (0.71) |
## |**Amyloid status** | | | |
## | Negative |373 (82%) |188 (84%) |185 (80%) |
## | Positive |83 (18%) |37 (16%) |46 (20%) |
## |**Amyloid SUVR** | | | |
## | Range |0.40, 1.27 |0.40, 1.25 |0.40, 1.27 |
## | Mean (SD) |0.59 (0.15) |0.59 (0.14) |0.60 (0.15) |
## |**Serum neurofilament light (pg/ml)** | | | |
## | Range |7.26, 124.00 |7.26, 121.00 |7.39, 124.00 |
## | Mean (SD) |20.74 (12.19) |20.95 (10.78) |20.53 (13.45) |
## |**APOE ε4 status** | | | |
## | Non-carrier |325 (71%) |165 (73%) |160 (69%) |
## | Carrier |131 (29%) |60 (27%) |71 (31%) |
## |**Alzheimer's Polygenic Risk Score (z-score)** | | | |
## | Number |426 |209 |217 |
## | Range |-3.15, 2.75 |-2.60, 2.75 |-3.15, 2.14 |
## | Mean (SD) |-0.06 (1.01) |-0.12 (0.98) |-0.01 (1.04) |
## |**rs1452628_T Allele number** | | | |
## | No copies |169 (40%) |88 (42%) |81 (38%) |
## | 1 copy |195 (46%) |95 (45%) |100 (47%) |
## | 2 copies |59 (14%) |27 (13%) |32 (15%) |
## |**Framingham Risk Score age 36** | | | |
## | Number |411 |203 |208 |
## | Range |0.58, 11.25 |0.58, 5.29 |1.69, 11.25 |
## | Mean (SD) |2.90 (1.74) |1.72 (0.79) |4.05 (1.65) |
## |**Framingham Risk Score age 69** | | | |
## | Number |443 |216 |227 |
## | Range |2.53, 68.75 |2.53, 62.95 |14.46, 68.75 |
## | Mean (SD) |25.90 (13.45) |16.66 (9.20) |34.69 (10.68) |
## |**Forced expiratory volume (L)** | | | |
## | Number |409 |208 |201 |
## | Range |0.37, 4.84 |0.84, 3.76 |0.37, 4.84 |
## | Mean (SD) |2.71 (0.68) |2.26 (0.42) |3.18 (0.56) |
## |**Grip strength (kg)** | | | |
## | Number |449 |220 |229 |
## | Range |11.00, 61.50 |11.00, 43.80 |14.80, 61.50 |
## | Mean (SD) |32.99 (10.90) |24.48 (5.59) |41.16 (8.16) |
## |**Walking speed (m/s)** | | | |
## | Number |431 |209 |222 |
## | Range |0.57, 2.22 |0.57, 2.22 |0.57, 1.99 |
## | Mean (SD) |1.08 (0.26) |1.06 (0.26) |1.11 (0.26) |
#### Create plots of age, BPA and sex differences
Summary of major brain disorders
kable(summary(as.factor(mydata5$Diagnosis)))| x | |
|---|---|
| 415 | |
| Alzheimer’s Disease | 2 |
| Alzheimer’s Disease (diagnosed by CL) | 1 |
| Bipolar Disorder | 2 |
| Brain metastasis | 1 |
| Clinical stroke | 4 |
| Clinical stroke and radiological cortical stroke | 1 |
| Clinical stroke and radiological cortical stroke | 2 |
| Depression requiring ECT | 2 |
| Epilepsy | 5 |
| Epilepsy | 1 |
| Hepatic encephalopathy prior to liver transplant (liver cirrhosis secondary to NASH) | 1 |
| Multiple Sclerosis | 1 |
| Multiple Sclerosis | 1 |
| Myotonic dystrophy | 1 |
| Parkinson’s Disease | 2 |
| Parkinson’s Disease & Epilepsy | 1 |
| Radiological cortical stroke | 9 |
| Radiological cortical stroke (haemorrhage) | 1 |
| Radiological MS | 1 |
| Subdurals requiring neurosurgery | 1 |
| TBI | 1 |
Relationship between age and brain predicted age
(sumstatsgraph <- ggarrange(age_density, sex_histogram,
ncol=1, nrow=2,
labels = "AUTO"))
4 Create model of all variables to check overall variance explained
overalllm <- lm(PAD ~ Agez + Sex + Childcog + Education + SocioStatus + Smoking + TIVz + WBVz + Ventvolz + Hippovolz + WMHVolz + PACC + AmyloidSuvrz + ApoE4 + PRS + rs1452628_T + SerumNFLz + FHS_36z + FHS_69z + MajorBrainDisorder + FEV1z + GripStrengthz + WalkSpeedz + Heightz, mydata5)
(summ(overalllm, confint = T))| Observations | 308 (148 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(27,280) | 6.15 |
| R² | 0.37 |
| Adj. R² | 0.31 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -0.96 | -4.94 | 3.02 | -0.47 | 0.64 |
| Agez | 0.84 | 0.02 | 1.66 | 2.01 | 0.05 |
| Sexmale | -0.46 | -4.29 | 3.37 | -0.24 | 0.81 |
| Childcog | 0.39 | -0.88 | 1.67 | 0.60 | 0.55 |
| EducationVocational or GCSE | -2.03 | -4.72 | 0.66 | -1.49 | 0.14 |
| EducationA-level or higher | -2.18 | -4.97 | 0.61 | -1.54 | 0.13 |
| SocioStatusNon-manual | 1.22 | -1.29 | 3.72 | 0.96 | 0.34 |
| SmokingEx-smoker | -1.05 | -2.80 | 0.69 | -1.19 | 0.24 |
| SmokingCurrent smoker | 0.64 | -4.80 | 6.08 | 0.23 | 0.82 |
| TIVz | 6.34 | 3.97 | 8.70 | 5.28 | 0.00 |
| WBVz | -3.90 | -5.93 | -1.87 | -3.78 | 0.00 |
| Ventvolz | -0.73 | -1.81 | 0.35 | -1.34 | 0.18 |
| Hippovolz | 0.92 | -0.10 | 1.95 | 1.77 | 0.08 |
| WMHVolz | 2.00 | 1.18 | 2.83 | 4.79 | 0.00 |
| PACC | -0.22 | -1.56 | 1.12 | -0.33 | 0.75 |
| AmyloidSuvrz | 0.23 | -0.68 | 1.14 | 0.50 | 0.61 |
| ApoE4Carrier | 0.06 | -1.85 | 1.96 | 0.06 | 0.95 |
| PRS | -0.40 | -1.19 | 0.38 | -1.01 | 0.31 |
| rs1452628_T1 | -1.15 | -2.89 | 0.58 | -1.31 | 0.19 |
| rs1452628_T2 | -0.46 | -3.01 | 2.08 | -0.36 | 0.72 |
| SerumNFLz | 1.33 | 0.57 | 2.10 | 3.44 | 0.00 |
| FHS_36z | 0.64 | -0.51 | 1.79 | 1.09 | 0.28 |
| FHS_69z | 0.21 | -1.01 | 1.44 | 0.34 | 0.73 |
| MajorBrainDisorderPresent | 2.40 | -0.49 | 5.29 | 1.64 | 0.10 |
| FEV1z | 0.27 | -0.99 | 1.54 | 0.43 | 0.67 |
| GripStrengthz | -0.10 | -1.40 | 1.19 | -0.16 | 0.88 |
| WalkSpeedz | -0.25 | -1.07 | 0.57 | -0.60 | 0.55 |
| Heightz | 0.35 | -1.13 | 1.82 | 0.46 | 0.64 |
| Standard errors: OLS |
5 Early lifecourse and demographic contributors to brain age
Chronological age
agelm <- lm(PAD ~ Age, data=mydata5)
summ(agelm, robust="HC1", digits = 1,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(1,454) | 6.5 |
| R² | 0.0 |
| Adj. R² | 0.0 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -103.1 | -175.9 | -30.2 | -2.8 | 0.0 |
| Age | 1.4 | 0.4 | 2.5 | 2.7 | 0.0 |
| Standard errors: Robust, type = HC1 |
plot(agelm) #check linearity assumptions
Sex
sexlm <- lm(PAD ~ Sex + Age, data=mydata5)
summ(sexlm, digits = 3,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(2,453) | 34.116 |
| R² | 0.131 |
| Adj. R² | 0.127 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -112.074 | -184.911 | -39.238 | -3.024 | 0.003 |
| Sexmale | 5.491 | 4.109 | 6.874 | 7.804 | 0.000 |
| Age | 1.507 | 0.477 | 2.538 | 2.875 | 0.004 |
| Standard errors: OLS |
plot(sexlm) #check linearity assumptions
coefssex <- tidy(sexlm , conf.int = TRUE) #modify estimate and confidence intervals for figure display
coefssex <- coefssex %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefssex$estimate, coefssex$conf.low, coefssex$conf.high )) %>%
dplyr::filter(term=="Sexmale")Educational attainment
educationlm <- lm(PAD ~ Education + Sex + Age, data=mydata5)
summ(educationlm, robust="HC1", digits = 1,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 17.3 |
| R² | 0.1 |
| Adj. R² | 0.1 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -113.2 | -181.5 | -45.0 | -3.3 | 0.0 |
| EducationVocational or GCSE | 1.1 | -1.2 | 3.3 | 0.9 | 0.3 |
| EducationA-level or higher | 0.9 | -1.2 | 2.9 | 0.8 | 0.4 |
| Sexmale | 5.5 | 4.1 | 6.9 | 7.6 | 0.0 |
| Age | 1.5 | 0.5 | 2.5 | 3.1 | 0.0 |
| Standard errors: Robust, type = HC1 |
plot(educationlm) #check linearity assumptions
coefsedu <- tidy(educationlm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsedu <- coefsedu %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsedu$estimate, coefsedu$conf.low, coefsedu$conf.high )) %>%
dplyr::filter(term=="EducationVocational or GCSE" | term== "EducationA-level or higher") Socioeconomic status
sociolm <- lm(PAD ~ SocioStatus + Sex + Age, data=mydata5)
summ(sociolm, robust="HC1", digits = 1,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 24.1 |
| R² | 0.1 |
| Adj. R² | 0.1 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -115.4 | -183.4 | -47.4 | -3.3 | 0.0 |
| SocioStatusNon-manual | 1.9 | -0.1 | 3.8 | 1.9 | 0.1 |
| Sexmale | 5.6 | 4.2 | 6.9 | 7.9 | 0.0 |
| Age | 1.5 | 0.6 | 2.5 | 3.1 | 0.0 |
| Standard errors: Robust, type = HC1 |
plot(sociolm) #check linearity assumptions
coefssocio <- tidy(sociolm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefssocio <- coefssocio %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefssocio$estimate, coefssocio$conf.low, coefssocio$conf.high )) %>%
dplyr::filter(term=="SocioStatusNon-manual")Childhood cognition (IQ)
childcoglm <- lm(PAD ~ Childcog + Sex + Age, data=mydata5)
summ(childcoglm, robust="HC1", digits = 1,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 23.7 |
| R² | 0.1 |
| Adj. R² | 0.1 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -108.5 | -176.0 | -41.1 | -3.2 | 0.0 |
| Childcog | 0.8 | -0.1 | 1.6 | 1.7 | 0.1 |
| Sexmale | 5.5 | 4.2 | 6.9 | 7.9 | 0.0 |
| Age | 1.5 | 0.5 | 2.4 | 3.0 | 0.0 |
| Standard errors: Robust, type = HC1 |
plot(childcoglm) #check linearity assumptions
coefschildcog <- tidy(childcoglm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefschildcog <- coefschildcog %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefschildcog$estimate, coefschildcog$conf.low, coefschildcog$conf.high )) %>%
dplyr::filter(term=="Childcog")Modify coeficients for graphical presentation
coefsel <- rbind(coefssex, coefschildcog, coefsedu, coefssocio)
# Add nicer names
coefsel$names <- factor(c( "Male", "Childhood IQ (z score)",
"Vocational or GCSE", "A level or higher", "Non-manual social class"),
levels = c("Male", "Childhood IQ (z score)",
"Vocational or GCSE", "A level or higher", "Non-manual social class"))
coefsel$type = factor(c("Sex",
"Childhood cognition",
rep("Education level (reference: none attempted)", 2),
"Socioeconomic Status (reference: manual social class)"),
levels = c("Sex", "Childhood cognition", "Education level (reference: none attempted)",
"Socioeconomic Status (reference: manual social class)"))6 Midlife contributions to brain age - cardiovascular risk (FHS) at 36 and 69
FHS at age 36
FHS36lm <- lm(PAD ~ FHS_36z + SocioStatus, data=mydata5)
summ(FHS36lm, robust="HC1", digits = 4,
confint=T)| Observations | 411 (45 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(2,408) | 19.1344 |
| R² | 0.0858 |
| Adj. R² | 0.0813 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.7732 | -5.6058 | -1.9407 | -4.0476 | 0.0001 |
| FHS_36z | 2.3554 | 1.5960 | 3.1148 | 6.0974 | 0.0000 |
| SocioStatusNon-manual | 1.3944 | -0.6218 | 3.4106 | 1.3596 | 0.1747 |
| Standard errors: Robust, type = HC1 |
plot(FHS36lm)
Scatterplot with FHS 36
FHS age 69
FHS69lm <- lm(PAD ~ FHS_69z + SocioStatus, data=mydata5)
summ(FHS69lm, robust="HC1", digits = 4,
confint=T)| Observations | 443 (13 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(2,440) | 28.0354 |
| R² | 0.1130 |
| Adj. R² | 0.1090 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -4.2313 | -6.0351 | -2.4275 | -4.6103 | 0.0000 |
| FHS_69z | 2.6297 | 1.9945 | 3.2649 | 8.1369 | 0.0000 |
| SocioStatusNon-manual | 1.7078 | -0.2519 | 3.6675 | 1.7128 | 0.0875 |
| Standard errors: Robust, type = HC1 |
plot(FHS69lm) # Check linearity assumptions
Scatter plot of FHS 69
Combining coefficients for presentation
coefs36 <- tidy(FHS36lm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefs36 <- coefs36 %>%
dplyr::filter(term=="FHS_36z")
coefs36 <- coefs36 %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefs36$estimate, coefs36$conf.low, coefs36$conf.high ))coefs69 <- tidy(FHS69lm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefs69 <- coefs69 %>%
dplyr::filter(term=="FHS_69z")
coefs69 <- coefs69 %>%
dplyr::mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefs69$estimate, coefs69$conf.low, coefs69$conf.high ))coefsFHS <- dplyr::union(coefs36, coefs69) #Create table of coefficients for graphical display
# Add nicer names
coefsFHS$names <- factor(c("FRS age 36 (z score)", "FRS age 69 (z score)"), levels = c("FRS age 69 (z score)", "FRS age 36 (z score)"))
# Group the variables
coefsFHS$type = rep("Cardiovascular Risk", 2)Compare early life and mid life factors contributing to brain age with a Williams test
earlymidcomparedata <- na.omit(mydata5, cols=seq_along(c(childcog, FHS36)))
dim(earlymidcomparedata)## [1] 305 47# Create stepwise correlations of the 3 variables
cor12 <- cor.test(earlymidcomparedata$PAD, earlymidcomparedata$FHS_36z)$estimate
cor13 <- cor.test(earlymidcomparedata$PAD, earlymidcomparedata$Childcog)$estimate
cor23 <- cor.test(earlymidcomparedata$Childcog, earlymidcomparedata$FHS_36z)$estimate
psych::r.test(n = 305, r12=cor12, r13=cor13, r23=cor23) #Apply the williams test## Correlation tests
## Call:[1] "r.test(n = 305 , r12 = 0.290733030424664 , r23 = -0.0474120507613991 , r13 = 0.0226692971450102 )"
## Test of difference between two correlated correlations
## t value 3.3 with probability < 0.000947 Contemporary correlate of brain age: Imaging, genetic/blood biomarker and physical factors
Imaging findings
Whole brain volume
imagingWBV <- lm(PAD ~ WBVz + TIVz + Sex + Age, data=mydata5)
summ(imagingWBV, robust="HC1", scale =T, digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 35.8368 |
| R² | 0.2412 |
| Adj. R² | 0.2345 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.1960 | -4.2892 | -2.1028 | -5.7453 | 0.0000 |
| WBVz | -3.5426 | -5.0164 | -2.0687 | -4.7236 | 0.0000 |
| TIVz | 6.3078 | 4.6152 | 8.0003 | 7.3242 | 0.0000 |
| Sexmale | 0.8362 | -0.8953 | 2.5677 | 0.9491 | 0.3431 |
| Age | 0.6118 | -0.0225 | 1.2460 | 1.8956 | 0.0586 |
| Standard errors: Robust, type = HC1; Continuous predictors are mean-centered and scaled by 1 s.d. |
#Check linearity assumptions
plot(imagingWBV)
coefswbv <- tidy(imagingWBV , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefswbv <- coefswbv %>%
dplyr::filter(term=="WBVz")
coefswbv <- coefswbv %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefswbv$estimate, coefswbv$conf.low, coefswbv$conf.high ))
Hippocampal volume
imaginghippo <- lm(PAD ~ Hippovolz + Sex + TIVz + Age, data=mydata5)
summ(imaginghippo, robust="HC1", scale =T, digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 28.6347 |
| R² | 0.2025 |
| Adj. R² | 0.1955 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.6006 | -4.7275 | -2.4736 | -6.2789 | 0.0000 |
| Hippovolz | 0.1046 | -0.7306 | 0.9397 | 0.2460 | 0.8058 |
| Sexmale | 1.6348 | -0.1601 | 3.4297 | 1.7899 | 0.0741 |
| TIVz | 2.8369 | 1.8085 | 3.8653 | 5.4213 | 0.0000 |
| Age | 0.8989 | 0.2631 | 1.5346 | 2.7786 | 0.0057 |
| Standard errors: Robust, type = HC1; Continuous predictors are mean-centered and scaled by 1 s.d. |
plot(imaginghippo) # check linearity assumptions
coefshippo <- tidy(imaginghippo , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefshippo <- coefshippo %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefshippo$estimate, coefshippo$conf.low, coefshippo$conf.high )) %>%
dplyr::filter(term=="Hippovolz")Ventricular volume
| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 28.8835 |
| R² | 0.2039 |
| Adj. R² | 0.1969 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.6218 | -4.7462 | -2.4974 | -6.3301 | 0.0000 |
| Ventvolz | 0.3520 | -0.5044 | 1.2084 | 0.8077 | 0.4197 |
| TIVz | 2.7190 | 1.6648 | 3.7732 | 5.0686 | 0.0000 |
| Sexmale | 1.6768 | -0.1097 | 3.4632 | 1.8446 | 0.0658 |
| Age | 0.8685 | 0.2408 | 1.4961 | 2.7193 | 0.0068 |
| Standard errors: Robust, type = HC1; Continuous predictors are mean-centered and scaled by 1 s.d. |
plot(imagingvent)
coefsvent <- tidy(imagingvent , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsvent <- coefsvent %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsvent$estimate, coefsvent$conf.low, coefsvent$conf.high )) %>%
dplyr::filter(term=="Ventvolz")Imaging WMH
imagingwmh <- lm(PAD ~ WMHVolz + TIVz + Sex + Age, data=mydata5)
summ(imagingwmh, robust="HC1", scale =T, digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 38.7523 |
| R² | 0.2558 |
| Adj. R² | 0.2492 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.9005 | -5.0098 | -2.7912 | -6.9103 | 0.0000 |
| WMHVolz | 1.8782 | 1.2843 | 2.4720 | 6.2154 | 0.0000 |
| TIVz | 2.6585 | 1.6830 | 3.6340 | 5.3558 | 0.0000 |
| Sexmale | 2.2269 | 0.4630 | 3.9908 | 2.4811 | 0.0135 |
| Age | 0.7279 | 0.1167 | 1.3390 | 2.3406 | 0.0197 |
| Standard errors: Robust, type = HC1; Continuous predictors are mean-centered and scaled by 1 s.d. |
plot(imagingwmh)
coefsWMH <- tidy(imagingwmh, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsWMH <- coefsWMH %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsWMH$estimate, coefsWMH$conf.low, coefsWMH$conf.high )) %>%
dplyr::filter(term=="WMHVolz")Combine results of brain imaging models for graphical display
coefsimaging <- rbind(coefswbv, coefshippo, coefsvent, coefsWMH)
# Add nicer names
coefsimaging$names <- factor(c( "Whole brain (z score)", "Hippocampi (z score)", "Ventricles (z score)", "WMH (z score)"), levels = c("WMH (z score)", "Ventricles (z score)", "Hippocampi (z score)", "Whole brain (z score)"))
# Group the variables
coefsimaging$type = rep("MR Imaging volumes", 4)WMH scatter plot 
Biomarkers
NFL
NFLlm <- lm(PAD ~ SerumNFLz + Sex + Age, data = mydata5)
summNFL <- summ(NFLlm, robust="HC1", digits = 4,
confint=T)
summNFL| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 27.5979 |
| R² | 0.1548 |
| Adj. R² | 0.1492 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -106.4076 | -174.2287 | -38.5864 | -3.0833 | 0.0022 |
| SerumNFLz | 1.2443 | 0.5912 | 1.8974 | 3.7441 | 0.0002 |
| Sexmale | 5.5320 | 4.1670 | 6.8969 | 7.9647 | 0.0000 |
| Age | 1.4268 | 0.4670 | 2.3867 | 2.9214 | 0.0037 |
| Standard errors: Robust, type = HC1 |
coefsNFL <- tidy(NFLlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsNFL <- coefsNFL %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsNFL$estimate, coefsNFL$conf.low, coefsNFL$conf.high )) %>%
dplyr::filter(term=="SerumNFLz")plot(NFLlm) #check linearity assumptions
#ggsave("Figures/NFLscatter75.png", bpa_NFL_scatter, height=10, width=18, units="cm" )NFL removing outliers
#Identidy NFL outliers by Tukey's Fences
NFLoutliercutoff <- summary(mydata5$SerumNFL)[["3rd Qu."]] + 1.5*IQR(mydata5$SerumNFL)
# Create subset of participants who are outliers to compare to remainder of cohort
NFLoutliers <- mydata5 %>%
filter(mydata5$SerumNFL > NFLoutliercutoff)Scatterplot with NFL
Fibrillar amyloid beta: Continuous (SUVR) and discrete (amyloid) amyloid beta measurement
Continuous amyloid measurement (SUVR) scatterplot
(bpa_amyloid_scatter <- ggplot(mydata5, aes(x = AmyloidSuvr, y = BPA)) +
geom_smooth(method = 'lm', colour = "black", size=0.4, fill = "grey75") +
geom_point( size = 1, alpha = 0.7) +
theme_bw() +
stat_cor(method = "pearson", label.x.npc = 0.68, label.y.npc = 0) +
labs(x = "Amyloid SUVR", y = "Brain Predicted Age (years)") +
theme_bw() +
theme(panel.grid.major = element_line(colour = "black", size = 0.08),
panel.grid.minor = element_blank(),
strip.background = element_blank(),
strip.text = element_text(size = 12),
axis.text = element_text(size = 12),
axis.title = element_text(size = 12),
legend.text = element_text(size = 12),
legend.title = element_text(size = 12)))
Discrete amyloid measurement: amyloid positive verse negative
(bpa_amyloid_box <- ggplot(mydata5, aes(x=AmyloidStatus, y=BPA)) +
geom_boxplot()) +
geom_jitter(size = 0.7, alpha=0.5, colour="grey30" )
Continuous amyloid regression model
suvrlm <- lm(PAD ~ AmyloidSuvrz + Sex + Age, data = mydata5)
summ(suvrlm, robust="HC1", digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 24.5233 |
| R² | 0.1400 |
| Adj. R² | 0.1343 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -109.7892 | -177.4040 | -42.1744 | -3.1910 | 0.0015 |
| AmyloidSuvrz | 0.7672 | 0.0517 | 1.4827 | 2.1072 | 0.0356 |
| Sexmale | 5.4078 | 4.0218 | 6.7938 | 7.6679 | 0.0000 |
| Age | 1.4756 | 0.5188 | 2.4323 | 3.0309 | 0.0026 |
| Standard errors: Robust, type = HC1 |
plot(suvrlm)
Discrete amyloid regression model
amyloidlm <- lm(PAD ~ AmyloidStatus + Sex + Age, data = mydata5)
summ(amyloidlm, robust="HC1", digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 23.0516 |
| R² | 0.1327 |
| Adj. R² | 0.1269 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -112.4930 | -180.3695 | -44.6165 | -3.2570 | 0.0012 |
| AmyloidStatusPositive | 0.8813 | -0.9487 | 2.7112 | 0.9464 | 0.3445 |
| Sexmale | 5.4609 | 4.0735 | 6.8484 | 7.7351 | 0.0000 |
| Age | 1.5112 | 0.5508 | 2.4716 | 3.0922 | 0.0021 |
| Standard errors: Robust, type = HC1 |
plot(amyloidlm)
coefssuvr <- tidy(suvrlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefssuvr <- coefssuvr %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefssuvr$estimate, coefssuvr$conf.low, coefssuvr$conf.high )) %>%
dplyr::filter(term=="AmyloidSuvrz")coefsbiomarker <- rbind(coefsNFL, coefssuvr)
# Add nicer names and variable types
coefsbiomarker$names <- c( "Serum NFL (z score)", "Florbetapir SUVR (z score)")
coefsbiomarker$type = rep("Blood and Imaging Biomarkers", 2)Genetics
APOE
apoe4lm <- lm(PAD ~ ApoE4, data = mydata5)
summ(apoe4lm, robust="HC1", digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(1,454) | 0.5595 |
| R² | 0.0012 |
| Adj. R² | -0.0010 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -2.9513 | -3.8179 | -2.0846 | -6.6924 | 0.0000 |
| ApoE4Carrier | 0.6226 | -1.0406 | 2.2858 | 0.7356 | 0.4623 |
| Standard errors: Robust, type = HC1 |
anovaapo <- anova(apoe4lm)
anovaapoplot(apoe4lm) # Check linearity assumptions
coefsapoe <- tidy(apoe4lm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsapoe <- coefsapoe %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsapoe$estimate, coefsapoe$conf.low, coefsapoe$conf.high )) %>%
dplyr::filter(term !="(Intercept)")Brain age SNP
snplm <- lm(PAD ~ rs1452628_T, data = mydata5)
summ(snplm, robust="HC1", digits = 4,
confint=T)| Observations | 423 (33 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(2,420) | 0.0863 |
| R² | 0.0004 |
| Adj. R² | -0.0043 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -2.6770 | -3.8654 | -1.4886 | -4.4279 | 0.0000 |
| rs1452628_T1 | -0.1522 | -1.8147 | 1.5102 | -0.1800 | 0.8573 |
| rs1452628_T2 | 0.3450 | -2.0848 | 2.7748 | 0.2791 | 0.7803 |
| Standard errors: Robust, type = HC1 |
anova(snplm)plot(snplm) #Check linearity assumptions
coefssnp <- tidy(snplm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefssnp <- coefssnp %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefssnp$estimate, coefssnp$conf.low, coefssnp$conf.high )) %>%
dplyr::filter(term !="(Intercept)")Polygenic risk score for Alzheimer’s Disease
PRSlm <- lm(PAD ~ PRS, data = mydata5)
summ(PRSlm, robust="HC1", digits = 4,
confint=T)| Observations | 426 (30 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(1,424) | 0.4445 |
| R² | 0.0010 |
| Adj. R² | -0.0013 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -2.8165 | -3.5886 | -2.0443 | -7.1694 | 0.0000 |
| PRS | -0.2584 | -1.0181 | 0.5014 | -0.6684 | 0.5043 |
| Standard errors: Robust, type = HC1 |
anova(PRSlm)plot(PRSlm) # Check linearity assumptions
coefsPRS <- tidy(PRSlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsPRS <- coefsPRS %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsPRS$estimate, coefsPRS$conf.low, coefsPRS$conf.high )) %>%
dplyr::filter(term !="(Intercept)")Combine genetic results for graphical display
coefsgenetics <- rbind(coefssnp, coefsPRS, coefsapoe)
coefsgenetics# Add nicer names and variable types
coefsgenetics$names <- c("rs1452628_T heterozygote", "rs1452628_T homozygote", "AD Polygenic risk score (z score)", "APOE ε4 carrier")
coefsgenetics$type = rep("Genetics", 4)Cognition
Cognition Scatterplot 
Covariates chosen as those that have shown to contribute to PACC
coglm <- lm(PAD ~ PACC + Sex + Education + Childcog + SocioStatus + Age, data = mydata5)
summ(coglm, robust="HC1", digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(7,448) | 11.2724 |
| R² | 0.1498 |
| Adj. R² | 0.1365 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -110.3641 | -178.3490 | -42.3793 | -3.1904 | 0.0015 |
| PACC | -1.1598 | -2.3012 | -0.0184 | -1.9970 | 0.0464 |
| Sexmale | 5.3030 | 3.8465 | 6.7595 | 7.1554 | 0.0000 |
| EducationVocational or GCSE | 0.4855 | -1.9460 | 2.9170 | 0.3924 | 0.6949 |
| EducationA-level or higher | 0.0460 | -2.4656 | 2.5575 | 0.0360 | 0.9713 |
| Childcog | 0.9909 | -0.0242 | 2.0060 | 1.9184 | 0.0557 |
| SocioStatusNon-manual | 1.8943 | -0.3444 | 4.1330 | 1.6629 | 0.0970 |
| Age | 1.4535 | 0.4932 | 2.4138 | 2.9745 | 0.0031 |
| Standard errors: Robust, type = HC1 |
anova(coglm)plot(coglm) #check linearity assumptions
coefscog <- tidy(coglm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefscog <- coefscog %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefscog$estimate, coefscog$conf.low, coefscog$conf.high )) %>%
dplyr::filter(term =="PACC")coefscog$names <- c( "PACC (z score)")
coefscog$type = rep("Cognition", 1)Comorbidities - major brain disorders
(bdisorderbox <- ggplot(mydata5, aes(x=MajorBrainDisorder, y=BPA)) +
geom_boxplot() +
geom_jitter(size = 0.7, alpha=0.5, colour="grey30" ) +
labs(x = "Major Brain Disorder", y = "Brain Predicted Age (years)") +
theme(panel.grid.major = element_line(colour = "black", size = 0.08),
panel.grid.minor = element_blank(),
strip.background = element_blank(),
strip.text = element_text(size = 12),
axis.text = element_text(size = 12),
axis.title = element_text(size = 12),
legend.text = element_text(size = 12),
legend.title = element_text(size = 12)))
bdisorderlm <- lm(PAD ~ MajorBrainDisorder + Age + Sex, data = mydata5)
summ(bdisorderlm, robust="HC1", digits = 4,
confit = T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,452) | 24.7215 |
| R² | 0.1410 |
| Adj. R² | 0.1353 |
| Est. | S.E. | t val. | p | |
|---|---|---|---|---|
| (Intercept) | -113.6355 | 34.6425 | -3.2802 | 0.0011 |
| MajorBrainDisorderPresent | 2.8157 | 1.2200 | 2.3079 | 0.0215 |
| Age | 1.5262 | 0.4903 | 3.1127 | 0.0020 |
| Sexmale | 5.4368 | 0.6999 | 7.7679 | 0.0000 |
| Standard errors: Robust, type = HC1 |
plot(bdisorderlm)
coefsdisorder <- tidy(bdisorderlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsdisorder <- coefsdisorder %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsdisorder$estimate, coefsdisorder$conf.low, coefsdisorder$conf.high )) %>%
dplyr::filter(term =="MajorBrainDisorderPresent")coefsdisorder$names <- c( "Major brain disorder present")
coefsdisorder$type = rep("Comorbidities", 1)Physical attributes regression
FEV1
FEV1lm <- lm(PAD ~ FEV1z + Smoking + Heightz +Age + Sex, data = mydata5)
summ(FEV1lm, robust="HC1", digits = 4,
confint=T)| Observations | 409 (47 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(6,402) | 10.1828 |
| R² | 0.1319 |
| Adj. R² | 0.1190 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -131.7897 | -204.5604 | -59.0190 | -3.5603 | 0.0004 |
| FEV1z | 0.2129 | -0.9914 | 1.4173 | 0.3475 | 0.7284 |
| SmokingEx-smoker | 0.2252 | -1.3832 | 1.8335 | 0.2753 | 0.7833 |
| SmokingCurrent smoker | 2.8979 | -0.2650 | 6.0608 | 1.8012 | 0.0724 |
| Heightz | 0.4906 | -0.7747 | 1.7559 | 0.7622 | 0.4464 |
| Age | 1.7911 | 0.7586 | 2.8236 | 3.4102 | 0.0007 |
| Sexmale | 4.2837 | 1.8558 | 6.7115 | 3.4686 | 0.0006 |
| Standard errors: Robust, type = HC1 |
plot(FEV1lm) # Check linearity assumptions
coefsFEV1 <- tidy(FEV1lm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsFEV1 <- coefsFEV1 %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsFEV1$estimate, coefsFEV1$conf.low, coefsFEV1$conf.high )) %>%
dplyr::filter(term == "FEV1z")Grip strength
griplm <- lm(PAD ~ GripStrengthz + Sex + Age, data = mydata5)
summ(griplm, robust="HC1", digits = 4,
confint=T)| Observations | 449 (7 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(3,445) | 21.9741 |
| R² | 0.1290 |
| Adj. R² | 0.1232 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -115.3076 | -184.2102 | -46.4051 | -3.2889 | 0.0011 |
| GripStrengthz | -0.2095 | -1.2726 | 0.8535 | -0.3873 | 0.6987 |
| Sexmale | 5.7337 | 3.6188 | 7.8486 | 5.3281 | 0.0000 |
| Age | 1.5517 | 0.5762 | 2.5273 | 3.1260 | 0.0019 |
| Standard errors: Robust, type = HC1 |
plot(griplm) # Check linearity assumptions
coefsgrip <- tidy(griplm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsgrip <- coefsgrip %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsgrip$estimate, coefsgrip$conf.low, coefsgrip$conf.high )) %>%
dplyr::filter(term == "GripStrengthz")Walking speed
walklm <- lm(PAD ~ WalkSpeedz + Heightz + Sex + Age, data = mydata5)
summ(walklm, robust="HC1", digits = 4,
confint=T)| Observations | 431 (25 missing obs. deleted) |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,426) | 16.5566 |
| R² | 0.1345 |
| Adj. R² | 0.1264 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -115.4963 | -187.4414 | -43.5512 | -3.1554 | 0.0017 |
| WalkSpeedz | -0.0729 | -0.7535 | 0.6077 | -0.2105 | 0.8334 |
| Heightz | 0.3201 | -0.6916 | 1.3318 | 0.6219 | 0.5344 |
| Sexmale | 5.1689 | 3.1121 | 7.2258 | 4.9396 | 0.0000 |
| Age | 1.5579 | 0.5383 | 2.5774 | 3.0034 | 0.0028 |
| Standard errors: Robust, type = HC1 |
plot(walklm) # Check linearity assumptions
coefswalk <- tidy(walklm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefswalk <- coefswalk %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefswalk$estimate, coefswalk$conf.low, coefswalk$conf.high )) %>%
dplyr::filter(term == "WalkSpeedz")Combine physical attributes for graphical display
coefsphysical <- rbind(coefsFEV1, coefsgrip, coefswalk)
# Add nicer names and variable types
coefsphysical$names <- c("FEV1 (z score)", "Grip strength (z score)", "Walking speed (z score)")
coefsphysical$type = rep("Physical metrics", 3)8 Combined Scatterplots and boxplots
# 9 Combined forest plots
Combine model results to allow for creation of overall forest plots.
fullcoefsassoc <- rbind(coefsbiomarker, coefsgenetics, coefscog, coefsdisorder, coefsphysical, coefsel, coefsFHS, coefsimaging)fullcoefsassoc$type <- factor(fullcoefsassoc$type, levels=c("Sex", "Childhood cognition", "Education level (reference: none attempted)", "Socioeconomic Status (reference: manual social class)", "Cardiovascular Risk", "Genetics", "Blood and Imaging Biomarkers", "Cognition", "Comorbidities", "Physical metrics", "MR Imaging volumes"))fullcoefsassoc$names <- fct_relevel(fullcoefsassoc$names, "A level or higher", "Florbetapir SUVR (z score)", "rs1452628_T homozygote", "rs1452628_T heterozygote", "WMH (z score)", "Hippocampi (z score)", "Ventricles (z score)", "Walking speed (z score)", "Grip strength (z score)") (fullassocgraph <- ggplot(fullcoefsassoc, aes(estimate, names )) +
geom_point(aes(), size=0.9)+
geom_errorbarh(aes(xmin = conf.low, xmax = conf.high), height = 0, size = 0.4)+
geom_text(aes(x = Inf, label = display_ests), vjust = 0.5, colour = "black", hjust = -0.1, size = 3.5) +
coord_cartesian(clip = "off") +
geom_vline(xintercept = 0, linetype = 2)+
theme_bw(base_size = 12)+
theme(plot.margin = unit(c(1,4,0.5,0.5), "cm")) +
theme(panel.border = element_rect(fill = NA, colour="black", size = 0.33)) +
labs(x = NULL,y = NULL, title = NULL)+
facet_col(~type, scale = "free_y", space = "free") +
theme(strip.background = element_rect(fill="grey85", colour="black"))) 
#ggsave("Figures/combinedgraph17.png", plot=fullassocgraph, height=7.5, width = 7.5)10 Supplementary Material
Appendix 1 - summary data of those in the sample compared to the whole Insight 46 cohort and those participants with complete data
Create summary table outline used in the appendix
summary1 <- #Summary outline for use in the appendix
list("Chronological age (years)" =
list("Number" = ~ length(na.omit(.data$Age)),
"Range" = ~ paste(frmt(range(.data$Age, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Age, na_rm=T, show_n = "never", denote_sd = "paren")),
"Brain predicted age (years)" =
list("Number" = ~ length(na.omit(.data$BPA)),
"Range" = ~ paste(frmt(range(.data$BPA, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$BPA, na_rm=T, show_n = "never", denote_sd = "paren")),
"Brain predicted age difference (years)" =
list("Number" = ~ length(na.omit(.data$PAD)),
"Range" = ~ paste(frmt(range(.data$PAD, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PAD, na_rm=T, show_n = "never", denote_sd = "paren")),
"Sex" =
list("Female" = ~ qwraps2::n_perc0(.data$Sex == "female", na_rm = T, show_symbol = T)),
"Socioeconomic status" =
list("Manual" = ~ qwraps2::n_perc0(.data$SocioStatus == "Manual", na_rm = T, show_symbol = T),
"Non-manual"= ~ qwraps2::n_perc0(.data$SocioStatus == "Non-manual", na_rm = T, show_symbol = T)),
"Educational attainment" =
list("None attempted" = ~ qwraps2::n_perc0(.data$Education == "None attempted", na_rm = T, show_symbol = T),
"School grade" = ~ qwraps2::n_perc0(.data$Education == "Vocational or GCSE", na_rm = T, show_symbol = T),
"Higher education" = ~ qwraps2::n_perc0(.data$Education == "A-level or higher", na_rm = T, show_symbol = T)),
"Childhood cognition (z-score)" =
list("Number" = ~ length(na.omit(.data$Childcog)),
"Range" = ~ paste(frmt(range(.data$Childcog, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Childcog, na_rm=T, show_n = "never", denote_sd = "paren")),
"Smoking" =
list("Never smoked" = ~ qwraps2::n_perc0(.data$Smoking == "Never smoked", na_rm = T, show_symbol = T),
"Ex-smoker" = ~ qwraps2::n_perc0(.data$Smoking == "Ex-smoker", na_rm = T, show_symbol = T),
"Current smoker" = ~ qwraps2::n_perc0(.data$Smoking == "Current smoker", na_rm = T, show_symbol = T)),
"Major brain disorder" =
list("None" = ~ qwraps2::n_perc0(.data$MajorBrainDisorder == "None", na_rm = T, show_symbol = T),
"Present"= ~ qwraps2::n_perc0(.data$MajorBrainDisorder == "Present", na_rm = T, show_symbol = T)),
"Total intracranial volume (ml)" =
list("Number" = ~ length(na.omit(.data$TIV)),
"Range" = ~ paste(frmt(range(.data$TIV, na.rm=T), digits=0), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$TIV, na_rm=T, show_n = "never", denote_sd = "paren", digits=0)),
"Whole brain volume (ml)" =
list("Number" = ~ length(na.omit(.data$WBV)),
"Range" = ~ paste(frmt(range(.data$WBV, na.rm=T), digits=0), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WBV, na_rm=T, show_n = "never", denote_sd = "paren", digits=0)),
"Ventricular volume (ml)" =
list("Number" = ~ length(na.omit(.data$Ventvol)),
"Range" = ~ paste(frmt(range(.data$Ventvol, na.rm=T), digits=2), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Ventvol, na_rm=T, show_n = "never", denote_sd = "paren", digits=2)),
"Hippocampal volume (ml)" =
list("Number" = ~ length(na.omit(.data$Hippovol)),
"Range" = ~ paste(frmt(range(.data$Hippovol, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$Hippovol, na_rm=T, show_n = "never", denote_sd = "paren")),
"White matter hyperintensity volume (ml)" =
list("Number" = ~ length(na.omit(.data$WMHVol)),
"Range" = ~ paste(frmt(range(.data$WMHVol, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WMHVol, na_rm=T, show_n = "never", denote_sd = "paren")),
"PACC score (z-score)" =
list("Number" = ~ length(na.omit(.data$PACC)),
"Range" = ~ paste(frmt(range(.data$PACC, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PACC, na_rm=T, show_n = "never", denote_sd = "paren")),
"Amyloid status" =
list("Negative" = ~ qwraps2::n_perc0(.data$AmyloidStatus == "Negative", na_rm = T, show_symbol = T),
"Positive"= ~ qwraps2::n_perc0(.data$AmyloidStatus == "Positive", na_rm = T, show_symbol = T)),
"Amyloid SUVR" =
list("Number" = ~ length(na.omit(.data$AmyloidSuvr)),
"Range" = ~ paste(frmt(range(.data$AmyloidSuvr, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$AmyloidSuvr, na_rm=T, show_n = "never", denote_sd = "paren")),
"Serum Neurofilament light (pg/ml)" =
list("Number" = ~ length(na.omit(.data$SerumNFL)),
"Range" = ~ paste(frmt(range(.data$SerumNFL, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$SerumNFL, na_rm=T, show_n = "never", denote_sd = "paren")),
"APOE ε4 status" =
list("Non-carrier"= ~ qwraps2::n_perc0(.data$ApoE4 == "Non-carrier", na_rm = T, show_symbol = T),
"Carrier" = ~ qwraps2::n_perc0(.data$ApoE4 == "Carrier", na_rm = T, show_symbol = T)),
"Alzheimer's Polygenic Risk Score (z-score)" =
list("Number" = ~ length(na.omit(.data$PRS)),
"Range" = ~ paste(frmt(range(.data$PRS, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$PRS, na_rm=T, show_n = "never", denote_sd = "paren")),
"rs1452628_T Allele number" =
list("No copies"= ~ qwraps2::n_perc0(.data$rs1452628_T == "0", na_rm = T, show_symbol = T),
"1 copy" = ~ qwraps2::n_perc0(.data$rs1452628_T == "1", na_rm = T, show_symbol = T),
"2 copies" = ~ qwraps2::n_perc0(.data$rs1452628_T == "2", na_rm = T, show_symbol = T)),
"Framingham Risk Score age 36" =
list("Number" = ~ length(na.omit(.data$FHS_36)),
"Range" = ~ paste(frmt(range(.data$FHS_36, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FHS_36, na_rm=T, show_n = "never", denote_sd = "paren")),
"Framingham Risk Score age 69" =
list("Number" = ~ length(na.omit(.data$FHS_69)),
"Range" = ~ paste(frmt(range(.data$FHS_69, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FHS_69, na_rm=T, show_n = "never", denote_sd = "paren")),
"FEV1 (L)" =
list("Number" = ~ length(na.omit(.data$FEV1)),
"Range" = ~ paste(frmt(range(.data$FEV1, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$FEV1, na_rm=T, show_n = "never", denote_sd = "paren")),
"Grip strength (kg)" =
list("Number" = ~ length(na.omit(.data$GripStrength)),
"Range" = ~ paste(frmt(range(.data$GripStrength, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$GripStrength, na_rm=T, show_n = "never", denote_sd = "paren")),
"Walking speed (m/s)" =
list("Number" = ~ length(na.omit(.data$WalkSpeed)),
"Range" = ~ paste(frmt(range(.data$WalkSpeed, na.rm=T)), collapse=', '),
"Mean (SD)" =~ qwraps2::mean_sd(.data$WalkSpeed, na_rm=T, show_n = "never", denote_sd = "paren")))allsummary <- summary_table(allcases, summary1)
mysummary <- summary_table(mydata5, summary1)
completesummary <- summary_table(completedata, summary1)
nfloutliersummary <- summary_table(NFLoutliers, summary1)comparisonsummary <- cbind(mysummary, completesummary, allsummary, nfloutliersummary)
print(comparisonsummary,
rtitle = "Summary Statistics",
cnames = c("Sample used in study", "Complete cases", "All Insight 46 participants", "sNFL outliers"))##
##
## |Summary Statistics |Sample used in study |Complete cases |All Insight 46 participants |sNFL outliers |
## |:----------------------------------------------|:--------------------|:--------------|:---------------------------|:-------------|
## |**Chronological age (years)** | | | | |
## | Number |456 |305 |502 |14 |
## | Range |69.25, 71.87 |69.25, 71.85 |69.25, 71.87 |69.57, 71.85 |
## | Mean (SD) |70.67 (0.67) |70.61 (0.68) |70.67 (0.67) |70.87 (0.79) |
## |**Brain predicted age (years)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |46.29, 94.26 |46.29, 94.26 |46.29, 94.26 |55.23, 91.16 |
## | Mean (SD) |67.90 (8.15) |67.79 (8.33) |67.96 (8.15) |74.08 (9.74) |
## |**Brain predicted age difference (years)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |-24.59, 22.70 |-24.59, 22.70 |-24.59, 22.70 |-16.62, 20.61 |
## | Mean (SD) |-2.77 (8.04) |-2.82 (8.21) |-2.70 (8.05) |3.21 (9.77) |
## |**Sex** | | | | |
## | Female |225 (49%) |152 (50%) |246 (49%) |5 (36%) |
## |**Socioeconomic status** | | | | |
## | Manual |70 (15%) |42 (14%) |76 (15%) |2 (14%) |
## | Non-manual |386 (85%) |263 (86%) |426 (85%) |12 (86%) |
## |**Educational attainment** | | | | |
## | None attempted |70 (15%) |45 (15%) |78 (16%) |3 (21%) |
## | School grade |139 (30%) |97 (32%) |151 (30%) |5 (36%) |
## | Higher education |247 (54%) |163 (53%) |273 (54%) |6 (43%) |
## |**Childhood cognition (z-score)** | | | | |
## | Number |456 |305 |502 |14 |
## | Range |-1.60, 2.50 |-1.60, 2.50 |-1.60, 2.50 |-1.31, 1.25 |
## | Mean (SD) |0.41 (0.75) |0.43 (0.73) |0.39 (0.74) |0.05 (0.76) |
## |**Smoking** | | | | |
## | Never smoked |160 (35%) |105 (34%) |171 (34%) |4 (29%) |
## | Ex-smoker |280 (61%) |191 (63%) |313 (62%) |10 (71%) |
## | Current smoker |16 (4%) |9 (3%) |18 (4%) |0 (0%) |
## |**Major brain disorder** | | | | |
## | None |415 (91%) |278 (91%) |460 (92%) |12 (86%) |
## | Present |41 (9%) |27 (9%) |42 (8%) |2 (14%) |
## |**Total intracranial volume (ml)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |1,114, 1,939 |1,153, 1,939 |1,114, 1,939 |1,289, 1,600 |
## | Mean (SD) |1,431 (133) |1,431 (131) |1,433 (133) |1,448 (95) |
## |**Whole brain volume (ml)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |819, 1,494 |865, 1,494 |819, 1,494 |969, 1,263 |
## | Mean (SD) |1,099 (99) |1,100 (99) |1,100 (99) |1,109 (101) |
## |**Ventricular volume (ml)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |6.16, 112.00 |6.50, 112.00 |6.16, 112.00 |18.43, 85.33 |
## | Mean (SD) |30.94 (16.34) |30.09 (15.69) |31.09 (16.33) |39.98 (20.85) |
## |**Hippocampal volume (ml)** | | | | |
## | Number |456 |305 |468 |14 |
## | Range |4.12, 8.54 |4.51, 8.54 |4.12, 8.54 |4.51, 7.81 |
## | Mean (SD) |6.26 (0.67) |6.29 (0.67) |6.26 (0.67) |6.10 (0.96) |
## |**White matter hyperintensity volume (ml)** | | | | |
## | Number |456 |305 |470 |14 |
## | Range |0.27, 33.67 |0.27, 33.67 |0.27, 33.67 |1.22, 24.90 |
## | Mean (SD) |5.21 (5.54) |5.20 (5.65) |5.21 (5.50) |6.95 (7.15) |
## |**PACC score (z-score)** | | | | |
## | Number |456 |305 |502 |14 |
## | Range |-3.49, 1.72 |-3.48, 1.72 |-3.49, 1.72 |-3.48, 1.58 |
## | Mean (SD) |-0.01 (0.74) |0.01 (0.72) |-0.00 (0.73) |-0.69 (1.30) |
## |**Amyloid status** | | | | |
## | Negative |373 (82%) |253 (83%) |378 (82%) |10 (71%) |
## | Positive |83 (18%) |52 (17%) |84 (18%) |4 (29%) |
## |**Amyloid SUVR** | | | | |
## | Number |456 |305 |462 |14 |
## | Range |0.40, 1.27 |0.40, 1.25 |0.40, 1.27 |0.53, 0.97 |
## | Mean (SD) |0.59 (0.15) |0.59 (0.14) |0.59 (0.15) |0.69 (0.15) |
## |**Serum Neurofilament light (pg/ml)** | | | | |
## | Number |456 |305 |500 |14 |
## | Range |7.26, 124.00 |7.26, 124.00 |7.21, 124.00 |38.60, 124.00 |
## | Mean (SD) |20.74 (12.19) |20.73 (12.97) |20.73 (11.87) |69.17 (34.03) |
## |**APOE ε4 status** | | | | |
## | Non-carrier |325 (71%) |219 (72%) |352 (70%) |10 (71%) |
## | Carrier |131 (29%) |86 (28%) |148 (30%) |4 (29%) |
## |**Alzheimer's Polygenic Risk Score (z-score)** | | | | |
## | Number |426 |305 |470 |13 |
## | Range |-3.15, 2.75 |-2.56, 2.60 |-3.15, 2.75 |-1.55, 2.75 |
## | Mean (SD) |-0.06 (1.01) |-0.03 (1.02) |-0.08 (1.00) |-0.21 (1.15) |
## |**rs1452628_T Allele number** | | | | |
## | No copies |169 (40%) |115 (38%) |183 (39%) |6 (43%) |
## | 1 copy |195 (46%) |152 (50%) |217 (47%) |6 (43%) |
## | 2 copies |59 (14%) |38 (12%) |65 (14%) |2 (14%) |
## |**Framingham Risk Score age 36** | | | | |
## | Number |411 |305 |455 |12 |
## | Range |0.58, 11.25 |0.58, 11.25 |0.58, 11.25 |1.57, 6.91 |
## | Mean (SD) |2.90 (1.74) |2.93 (1.82) |2.92 (1.78) |3.56 (1.82) |
## |**Framingham Risk Score age 69** | | | | |
## | Number |443 |305 |488 |13 |
## | Range |2.53, 68.75 |2.53, 68.75 |2.53, 83.87 |11.67, 50.97 |
## | Mean (SD) |25.90 (13.45) |25.83 (13.38) |26.16 (13.73) |27.97 (10.58) |
## |**FEV1 (L)** | | | | |
## | Number |409 |305 |450 |14 |
## | Range |0.37, 4.84 |0.37, 4.84 |0.37, 4.84 |0.84, 4.20 |
## | Mean (SD) |2.71 (0.68) |2.72 (0.67) |2.72 (0.68) |2.73 (0.82) |
## |**Grip strength (kg)** | | | | |
## | Number |449 |305 |494 |14 |
## | Range |11.00, 61.50 |13.20, 58.70 |11.00, 61.50 |17.20, 58.60 |
## | Mean (SD) |32.99 (10.90) |33.37 (11.04) |33.17 (10.93) |35.54 (13.42) |
## |**Walking speed (m/s)** | | | | |
## | Number |431 |305 |475 |13 |
## | Range |0.57, 2.22 |0.62, 2.22 |0.27, 2.22 |0.67, 1.84 |
## | Mean (SD) |1.08 (0.26) |1.10 (0.26) |1.08 (0.26) |1.00 (0.29) |Sensitivity analysis
Sex
agelesssexlm <- lm(PAD ~ Sex, data=mydata5)
coefsagelesssex <- tidy(agelesssexlm , conf.int = TRUE) #modify estimate and confidence intervals for figure display
coefsagelesssex <- coefsagelesssex %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesssex$estimate, coefsagelesssex$conf.low, coefsagelesssex$conf.high )) %>%
dplyr::filter(term=="Sexmale")Educational attainment
agelesseducationlm <- lm(PAD ~ Education + Sex, data=mydata5)
coefsagelessedu <- tidy(agelesseducationlm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessedu <- coefsagelessedu %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessedu$estimate, coefsagelessedu$conf.low, coefsagelessedu$conf.high )) %>%
dplyr::filter(term=="EducationVocational or GCSE" | term== "EducationA-level or higher") Socioeconomic status
agelesssociolm <- lm(PAD ~ SocioStatus + Sex, data=mydata5)
coefsagelesssocio <- tidy(agelesssociolm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesssocio <- coefsagelesssocio %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesssocio$estimate, coefsagelesssocio$conf.low, coefsagelesssocio$conf.high )) %>%
dplyr::filter(term=="SocioStatusNon-manual")Childhood cognition (IQ)
agelesschildcoglm <- lm(PAD ~ Childcog + Sex, data=mydata5)
summ(agelesschildcoglm, robust="HC1", digits = 1,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(2,453) | 31.2 |
| R² | 0.1 |
| Adj. R² | 0.1 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -5.9 | -7.0 | -4.8 | -10.8 | 0.0 |
| Childcog | 0.8 | -0.1 | 1.7 | 1.8 | 0.1 |
| Sexmale | 5.5 | 4.1 | 6.9 | 7.8 | 0.0 |
| Standard errors: Robust, type = HC1 |
coefsagelesschildcog <- tidy(agelesschildcoglm , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesschildcog <- coefsagelesschildcog %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesschildcog$estimate, coefsagelesschildcog$conf.low, coefsagelesschildcog$conf.high )) %>%
dplyr::filter(term=="Childcog")Modify coeficients for graphical presentation
coefsagelessel <- rbind(coefsagelesssex, coefsagelesschildcog, coefsagelessedu, coefsagelesssocio)
# Add nicer names
coefsagelessel$names <- factor(c( "Male", "Childhood IQ (z score)",
"Vocational or GCSE", "A level or higher", "Non-manual social class"),
levels = c("Male", "Childhood IQ (z score)",
"Vocational or GCSE", "A level or higher", "Non-manual social class"))
coefsagelessel$type = factor(c("Sex",
"Childhood cognition",
rep("Education level (reference: none attempted)", 2),
"Socioeconomic Status (reference: manual social class)"),
levels = c("Sex", "Childhood cognition", "Education level (reference: none attempted)",
"Socioeconomic Status (reference: manual social class)"))Ageless contemporary correlates
agelessimagingWBV <- lm(PAD ~ WBVz + TIVz + Sex, data=mydata5)
coefsagelesswbv <- tidy(agelessimagingWBV , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesswbv <- coefsagelesswbv %>%
dplyr::filter(term=="WBVz")
coefsagelesswbv <- coefsagelesswbv %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesswbv$estimate, coefsagelesswbv$conf.low, coefsagelesswbv$conf.high ))Hippocampi
agelessimaginghippo <- lm(PAD ~ Hippovolz + Sex + TIVz, data=mydata5)
coefsagelesshippo <- tidy(agelessimaginghippo , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesshippo <- coefsagelesshippo %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesshippo$estimate, coefsagelesshippo$conf.low, coefsagelesshippo$conf.high )) %>%
dplyr::filter(term=="Hippovolz")Imaging ventricles
agelessimagingvent <- lm(PAD ~ Ventvolz + TIVz + Sex, data=mydata5)
summ(imagingvent, robust="HC1", scale =T, digits = 4,
confint=T)| Observations | 456 |
| Dependent variable | PAD |
| Type | OLS linear regression |
| F(4,451) | 28.8835 |
| R² | 0.2039 |
| Adj. R² | 0.1969 |
| Est. | 2.5% | 97.5% | t val. | p | |
|---|---|---|---|---|---|
| (Intercept) | -3.6218 | -4.7462 | -2.4974 | -6.3301 | 0.0000 |
| Ventvolz | 0.3520 | -0.5044 | 1.2084 | 0.8077 | 0.4197 |
| TIVz | 2.7190 | 1.6648 | 3.7732 | 5.0686 | 0.0000 |
| Sexmale | 1.6768 | -0.1097 | 3.4632 | 1.8446 | 0.0658 |
| Age | 0.8685 | 0.2408 | 1.4961 | 2.7193 | 0.0068 |
| Standard errors: Robust, type = HC1; Continuous predictors are mean-centered and scaled by 1 s.d. |
coefsagelessvent <- tidy(agelessimagingvent , conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessvent <- coefsagelessvent %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessvent$estimate, coefsagelessvent$conf.low, coefsagelessvent$conf.high )) %>%
dplyr::filter(term=="Ventvolz")Imaging WMH
agelessimagingwmh <- lm(PAD ~ WMHVolz + TIVz + Sex, data=mydata5)
coefsagelessWMH <- tidy(agelessimagingwmh, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessWMH <- coefsagelessWMH %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessWMH$estimate, coefsagelessWMH$conf.low, coefsagelessWMH$conf.high )) %>%
dplyr::filter(term=="WMHVolz")Combine results of brain imaging models for graphical display
coefsagelessimaging <- rbind(coefsagelesswbv, coefsagelesshippo, coefsagelessvent, coefsagelessWMH)
# Add nicer names
coefsagelessimaging$names <- factor(c( "Whole brain (z score)", "Hippocampi (z score)", "Ventricles (z score)", "WMH (z score)"), levels = c("WMH (z score)", "Ventricles (z score)", "Hippocampi (z score)", "Whole brain (z score)"))
# Group the variables
coefsagelessimaging$type = rep("MR Imaging volumes", 4)agelessNFLlm <- lm(PAD ~ SerumNFLz + Sex + Age, data = mydata5)
coefsagelessNFL <- tidy(agelessNFLlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessNFL <- coefsagelessNFL %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessNFL$estimate, coefsagelessNFL$conf.low, coefsagelessNFL$conf.high )) %>%
dplyr::filter(term=="SerumNFLz")agelesssuvrlm <- lm(PAD ~ AmyloidSuvrz + Sex, data = mydata5)
coefsagelesssuvr <- tidy(agelesssuvrlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesssuvr <- coefsagelesssuvr %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesssuvr$estimate, coefsagelesssuvr$conf.low, coefsagelesssuvr$conf.high )) %>%
dplyr::filter(term=="AmyloidSuvrz")coefsagelessbiomarker <- rbind(coefsagelessNFL, coefsagelesssuvr)
coefsagelessbiomarker# Add nicer names and variable types
coefsagelessbiomarker$names <- c( "Serum NFL (z score)", "Florbetapir SUVR (z score)")
coefsagelessbiomarker$type = rep("Blood and Imaging Biomarkers", 2)Cognition
agelesscoglm <- lm(PAD ~ PACC + Sex + Education + Childcog + SocioStatus, data = mydata5)
coefsagelesscog <- tidy(agelesscoglm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesscog <- coefsagelesscog %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesscog$estimate, coefsagelesscog$conf.low, coefsagelesscog$conf.high )) %>%
dplyr::filter(term =="PACC")
coefsagelesscog$names <- c( "PACC (z score)")
coefsagelesscog$type = rep("Cognition", 1)agelessdisorderlm <- lm(PAD ~ MajorBrainDisorder + Sex, data = mydata5)
coefsagelessdisorder <- tidy(agelessdisorderlm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessdisorder <- coefsagelessdisorder %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessdisorder$estimate, coefsagelessdisorder$conf.low, coefsagelessdisorder$conf.high )) %>%
dplyr::filter(term =="MajorBrainDisorderPresent")
coefsagelessdisorder$names <- c( "Major Brain Disorder")
coefsagelessdisorder$type = rep("Comorbidities", 1)Physical attributes regression
agelessFEV1lm <- lm(PAD ~ FEV1z + Smoking + Heightz + Sex, data = mydata5)
coefsagelessFEV1 <- tidy(agelessFEV1lm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessFEV1 <- coefsagelessFEV1 %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessFEV1$estimate, coefsagelessFEV1$conf.low, coefsagelessFEV1$conf.high )) %>%
dplyr::filter(term == "FEV1z")agelessgriplm <- lm(PAD ~ GripStrengthz + Sex, data = mydata5)
coefsagelessgrip <- tidy(agelessgriplm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelessgrip <- coefsagelessgrip %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelessgrip$estimate, coefsagelessgrip$conf.low, coefsagelessgrip$conf.high )) %>%
dplyr::filter(term == "GripStrengthz")agelesswalklm <- lm(PAD ~ WalkSpeedz + Heightz + Sex, data = mydata5)
coefsagelesswalk <- tidy(agelesswalklm, conf.int = TRUE) #modify estimate and confidence intervals for graphical display
coefsagelesswalk <- coefsagelesswalk %>%
mutate(display_ests = sprintf("%.1f [%.1f, %.1f]" , coefsagelesswalk$estimate, coefsagelesswalk$conf.low, coefsagelesswalk$conf.high )) %>%
dplyr::filter(term == "WalkSpeedz")
coefsagelessphysical <- rbind(coefsagelessFEV1, coefsagelessgrip, coefsagelesswalk)
# Add nicer names and variable types
coefsagelessphysical$names <- c("FEV1 (z score)", "Grip strength (z score)", "Walking speed (z score)")
coefsagelessphysical$type = rep("Physical metrics", 3)Combined forest plots
fullagelesscoefsassoc <- rbind(coefsagelessbiomarker, coefsagelesscog, coefsagelessdisorder, coefsagelessphysical, coefsagelessel, coefsagelessimaging)fullagelesscoefsassoc$type <- factor(fullagelesscoefsassoc$type, levels=c("Sex", "Childhood cognition", "Education level (reference: none attempted)", "Socioeconomic Status (reference: manual social class)", "Cardiovascular Risk", "Genetics", "Blood and Imaging Biomarkers", "Cognition", "Comorbidities", "Physical metrics", "MR Imaging volumes"))fullagelesscoefsassoc$names <- fct_relevel(fullagelesscoefsassoc$names, "A level or higher", "Florbetapir SUVR (z score)", "WMH (z score)", "Hippocampi (z score)", "Ventricles (z score)", "Walking speed (z score)", "Grip strength (z score)") (fullagelessassocgraph <- ggplot(fullagelesscoefsassoc, aes(estimate, names )) +
geom_point(aes(), size=0.9)+
geom_errorbarh(aes(xmin = conf.low, xmax = conf.high), height = 0, size = 0.4)+
geom_text(aes(x = Inf, label = display_ests), vjust = 0.5, colour = "black", hjust = -0.1, size = 3.5) +
coord_cartesian(clip = "off") +
geom_vline(xintercept = 0, linetype = 2)+
theme_bw(base_size = 12)+
theme(plot.margin = unit(c(1,4,0.5,0.5), "cm")) +
theme(panel.border = element_rect(fill = NA, colour="black", size = 0.33)) +
labs(x = NULL,y = NULL, title = NULL)+
facet_col(~type, scale = "free_y", space = "free") +
theme(strip.background = element_rect(fill="grey85", colour="black"))) 
11 Version control
print(version)## _
## platform x86_64-apple-darwin17.0
## arch x86_64
## os darwin17.0
## system x86_64, darwin17.0
## status
## major 4
## minor 0.3
## year 2020
## month 10
## day 10
## svn rev 79318
## language R
## version.string R version 4.0.3 (2020-10-10)
## nickname Bunny-Wunnies Freak Outprint(sessionInfo())## R version 4.0.3 (2020-10-10)
## Platform: x86_64-apple-darwin17.0 (64-bit)
## Running under: macOS Catalina 10.15.7
##
## Matrix products: default
## BLAS: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRblas.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib
##
## locale:
## [1] en_AU.UTF-8/en_AU.UTF-8/en_AU.UTF-8/C/en_AU.UTF-8/en_AU.UTF-8
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] ggforce_0.3.2 jtools_2.1.2 skimr_2.1.2 broom_0.7.4
## [5] ggpubr_0.4.0 forcats_0.5.1 stringr_1.4.0 dplyr_1.0.3
## [9] purrr_0.3.4 readr_1.4.0 tidyr_1.1.2 tibble_3.0.6
## [13] ggplot2_3.3.3 tidyverse_1.3.0 qwraps2_0.5.0 psych_2.0.12
## [17] kableExtra_1.3.1 ggExtra_0.9
##
## loaded via a namespace (and not attached):
## [1] nlme_3.1-151 fs_1.5.0 lubridate_1.7.9.2 webshot_0.5.2
## [5] httr_1.4.2 repr_1.1.3 tools_4.0.3 backports_1.2.1
## [9] R6_2.5.0 mgcv_1.8-33 DBI_1.1.1 colorspace_2.0-0
## [13] withr_2.4.1 tidyselect_1.1.0 mnormt_2.0.2 curl_4.3
## [17] compiler_4.0.3 cli_2.2.0 rvest_0.3.6 xml2_1.3.2
## [21] sandwich_3.0-0 labeling_0.4.2 scales_1.1.1 digest_0.6.27
## [25] foreign_0.8-81 rmarkdown_2.6 rio_0.5.16 base64enc_0.1-3
## [29] pkgconfig_2.0.3 htmltools_0.5.1.1 highr_0.8 dbplyr_2.0.0
## [33] fastmap_1.1.0 rlang_0.4.10 readxl_1.3.1 rstudioapi_0.13
## [37] shiny_1.6.0 farver_2.0.3 generics_0.1.0 zoo_1.8-8
## [41] jsonlite_1.7.2 zip_2.1.1 car_3.0-10 magrittr_2.0.1
## [45] Matrix_1.3-2 Rcpp_1.0.6 munsell_0.5.0 fansi_0.4.2
## [49] abind_1.4-5 lifecycle_0.2.0 stringi_1.5.3 yaml_2.2.1
## [53] carData_3.0-4 MASS_7.3-53 grid_4.0.3 parallel_4.0.3
## [57] promises_1.1.1 crayon_1.4.0 miniUI_0.1.1.1 lattice_0.20-41
## [61] cowplot_1.1.1 splines_4.0.3 haven_2.3.1 pander_0.6.3
## [65] hms_1.0.0 tmvnsim_1.0-2 knitr_1.31 pillar_1.4.7
## [69] ggsignif_0.6.0 reprex_1.0.0 glue_1.4.2 evaluate_0.14
## [73] data.table_1.13.6 modelr_0.1.8 tweenr_1.0.1 vctrs_0.3.6
## [77] httpuv_1.5.5 cellranger_1.1.0 polyclip_1.10-0 gtable_0.3.0
## [81] assertthat_0.2.1 xfun_0.20 openxlsx_4.2.3 mime_0.9
## [85] xtable_1.8-4 rstatix_0.6.0 later_1.1.0.1 viridisLite_0.3.0
## [89] ellipsis_0.3.1